Choosing help: an AI consultant or an agency

An AI consultant or an agency: what each is structured to sell

Hire an AI consultant when the problem needs senior judgment applied deeply to one operation, and an AI agency when the work needs many hands moving in parallel. That is the structural difference, and every rate question that travels with this search - hourly rate, price, implementation cost - follows from it, because each option’s price is built from what it is structured to sell: a consultant prices scarce senior time, an agency prices a coordinated team. One disclosure before the comparison: Foxnut Studios is a contestant on this page. It is neither an independent consultant nor an agency but a two-person senior studio, a third structure with its own failure mode, and this comparison names the situations where its own category loses - starting with the agency’s genuine edge. The line the studio draws for itself runs through what this studio actually does in AI.

The three structures, compared on what changes the outcome

The axes below are the ones that move a buyer’s result: what is actually being purchased, what the price is built from, who touches the work, and what remains when the engagement ends.

What AI consultant rates actually price

An AI consultant’s hourly rate is the least informative number in this decision, because an hourly rate prices attendance and the thing being bought is a working system. For AI implementation work the honest unit is the scoped outcome: what gets built, proven on real work, documented, and handed to a team trained to run it. Two vendors quoting the same hourly number can differ several-fold in what a finished system costs, which is why the rate tables published around this search mostly describe other people’s engagements and settle nothing about yours.

What a buyer can compare across structures is what each price is built from. An agency’s quote carries the delivery team, the coordination between its workstreams, the account layer and the margin on all three - that is not padding, it is the cost of the volume the agency exists to deliver. An independent consultant’s quote carries scarce senior time and little else, which is why it looks cheap until the work needs more than one pair of hands. A senior studio’s quote carries seniority and scope: the cost of putting its principals on the work, attached to a defined outcome. Foxnut Studios publishes its own numbers rather than gesturing at a market: the published ranges for a Foxnut Studios engagement are stated in plain text, by engagement shape, on one page site-wide. And when a budget and a brief do not match, nobody is sold a shrunk version of the scope - the studio offers a free audit call instead and points at the best alternatives.

What an AI consultant actually is

The title is unregulated, which is most of the buyer’s problem. As this library uses the term, an AI consultant is a practitioner who has built and shipped production AI systems first-hand and now applies that record to your operation - judgment with proof behind it, not a framework with a fee attached. Anything less is a strategy document with a consultant’s signature, and the market currently carries many of those. The test is retrospective: has this person done, in production, the specific thing you are hiring them to guide? How to run that test in a first conversation is a deep enough subject to have its own page in this territory.

When to choose each

Each structure wins a real category of problem and fails in a characteristic way. An option whose failure mode nobody will name has not been compared, only advertised.

When an independent AI consultant wins

An independent consultant wins when the problem is one discipline deep and judgment is the whole purchase: an architecture decision, a model and tooling call, a diagnosis of why the pilot stalled. One senior person, close to the problem, no coordination overhead. The failure mode is coverage, and it is sharper in AI than elsewhere: a production build spans data plumbing, prompt and model work, evaluation, deployment and training the client’s team, and one person is rarely deep in all of it. Add the ordinary fragility of a single calendar - one illness, one better-paying client - and the engagement is what slips.

When an AI agency wins

An agency wins on execution volume, and the concession deserves to be made plainly rather than buried: when the brief is ten workflows across three departments this quarter, integrations that need maintaining, and delivery that must survive any one person’s departure or holiday, the agency model is not a compromise - it is the only structure on this page built for that load. Staffing depth, parallel workstreams and process are real assets, and a two-person studio cannot match them and should say so. The failure mode is the bench. An agency’s depth is an average across its roster: the senior people who scoped your engagement move to the next sale, delivery lands with whoever is available, and an account layer translates between you and the people doing the work - while the rate stays senior even when the depth assigned to you is not. For deterministic execution at volume, the model works as designed. For work that needs judgment on every step, the average is what you get, and the exceptional agencies whose average runs senior are genuinely rare and priced accordingly.

When a small senior studio wins

A senior studio wins when the problem needs both the judgment and the build, and you want the people who scoped the work to be the people who do it. A team of two fields its principals on every engagement because there is nobody else to hand the work down to - depth per person is the entire model, and what it buys is that the thinking and the building never separate. The failure mode is capacity, stated up front: a studio this size cannot absorb a doubling of scope mid-engagement, takes deliberately few engagements at once, and the only honest response at the ceiling is no. The refusal list on this page is what that looks like in practice.

The question under the rate question

Ask any candidate - consultant, agency or studio - three things in writing before comparing a single rate: who exactly will do the work, what the price is built from, and what your team will own and run when they leave. The rate conversation mostly resolves itself once those are answered, because a senior day rate on the person actually delivering is cheap next to a discounted team you never meet, and expensive next to an off-the-shelf tool when the problem never needed deep expertise at all. What a finished handover should physically contain, and what happens to scope when a pilot succeeds, are each deep enough subjects to have their own pages in this territory - and they are the questions the rate pages skip.

The part most pages leave out

When not to choose Foxnut Studios

Situations where another option is the better call, and where we say so in the first conversation rather than the fourth.

• The client wants a superficial AI solution - something that could be bought off the shelf or configured in an afternoon, and does not require the studio's expertise. An off-the-shelf tool, or a generalist implementer. If the problem does not need deep expertise, paying for deep expertise is the wrong purchase - the studio says so and points at the simpler option.
• The client's senior leadership does not have conviction in making AI work for the organisation. Nobody, yet. Without conviction and commitment from the top, these projects tend to fail despite the best implementation - the honest move is to decline until leadership is committed, not to build something that will be abandoned.
• Some of the organisation's data is offline or simply undocumented. The organisation itself, doing the documentation work first. Data that is unstructured and scattered across many tools and databases is workable - the studio works with that routinely. Data that exists only in someone's head or on paper is not, and no consultant can fix that from outside.
• The brief needs work that neither the founders nor their expert network has done first-hand. A practitioner who has shipped that exact work. The filter is the studio's founding rule, and the network extends it rather than repeals it: any brief the studio takes is worked by people who have done the thing before.
• Taking the brief would put the studio in a conflict of interest with an existing client's work. A firm with no stake in the outcome. The studio says which conflict exists to whatever degree it can without breaching the first client's confidence, and declines.

Foxnut Studios works on briefs like this one from Bengaluru and Paris. If you want the shape of that before you talk to anyone, here is what an AI engagement covers and what you keep.